The story begins with a familiar modern scene: a piece of enormous news breaks, people reach for their phones, and social media fills the gap before anyone has had time to verify what is real. According to the article, just hours after news spread online that Nicolás Maduro, Venezuela’s ousted president, had been arrested by U.S. forces, feeds were flooded with photographs that appeared to show him in handcuffs, escorted by drug enforcement agents, or surrounded by soldiers on a military aircraft. The images looked convincing enough to travel quickly, but they were fake—likely produced with artificial intelligence tools. Experts described the episode as one of the first times A.I. imagery had depicted prominent figures while a historical moment was still rapidly unfolding. “This was the first time I’d personally seen so many A.I.-generated images of what was supposed to be a real moment in time,” said Roberta Braga, executive director of the Digital Democracy Institute of the Americas, a think tank. The unsettling part is that this was not some obscure corner of the internet. Most major A.I. image generators have rules against deceptive or misleading content, and several explicitly ban fake images of public figures like Maduro. Yet when The New York Times tested a dozen tools on Monday, many mainstream generators were willing to create similar images, often free of charge and in a matter of seconds. The long-feared threat that such fakes could confuse and mislead people as news breaks has now become a reality, because the technology has become both advanced enough and popular enough to make it happen.
The speed of the spread was part of the problem. Jeanfreddy Gutiérrez, who runs a fact-checking operation in Caracas but is based in Colombia, noticed the fake images of Maduro circulating on Saturday, the day his arrest was announced. He said some of them moved through Latin American news outlets before being quietly swapped out for an image shared by President Trump. Over the weekend, Gutiérrez published a fact-check identifying several deepfake images of Maduro, including one that “showed clear signs” it had been made by Gemini, Google’s A.I. app. “They spread so fast—I saw them in almost every Facebook and WhatsApp contact I have,” he said. NewsGuard, a company that tracks the reliability of online information, counted five fabricated or out-of-context images and two misrepresented videos linked to Maduro’s capture. Together, that content drew more than 14.1 million views on X in under two days, according to the company, and it also appeared with more limited engagement on Meta platforms such as Instagram and Facebook. What makes this especially troubling is that the fakes arrived before a clear, authoritative visual record had been established. In that vacuum, people were not simply sharing news; they were sharing their expectations, fears, and political instincts dressed up as photographs. The fake images became a kind of instant folklore for a moment that was still being written, and the fact that some news outlets briefly circulated them shows how easily even professional gatekeepers can be outpaced when the tools are fast, cheap, and convincing.
The Times’s tests exposed how uneven the safeguards are. After Maduro’s arrest, the newspaper tried a dozen A.I. generators to see which ones would create fake images of him. Most of the tools, including Gemini and models from OpenAI and X, quickly produced the requested images. That undercut assurances from many technology companies that they have designed safeguards to prevent abuse. Google has rules prohibiting users from generating misinformation and specifically bars misleading content “related to governmental or democratic processes.” But a Google spokesman said the company does not categorically bar images of prominent people, and a fake image depicting Maduro being arrested that was generated in the Times’s tests did not violate its rules. The spokesman added that Google has focused on a hidden watermark called SynthID, which Gemini embeds in every image it creates, enabling people to determine whether an image was made by its A.I. tool. “You can simply upload an image to the Gemini app and instantly confirm whether it was generated using Google A.I.,” he said. OpenAI’s tool, ChatGPT, said it could not create images of Maduro. But when the Times queried through another website that taps into the same model of ChatGPT, it relented and produced the images. An OpenAI spokesperson said by email that the company uses safeguards to protect public figures but focuses on preventing harms such as sexual deepfakes or violence. Grok, the model by X.ai, immediately produced lifelike images of Maduro’s arrest. Grok has already come under heavy scrutiny, including last week when watchdogs reported that it was responding to requests to remove clothing from images of children. Grok’s account posted that its team had identified “lapses in safeguards” and was “urgently fixing them.” X.ai did not respond to a request for comment. The pattern is not that every tool fails in exactly the same way, but that the guardrails are inconsistent, sometimes easy to bypass, and often focused on narrow categories of harm rather than the broader problem of political deception.
Other tools behaved differently, but not always in reassuring ways. Facebook’s Meta A.I. chatbot and Flux.ai responded by creating images showing a man with a mustache being arrested, but they did not accurately depict Maduro’s features. When prompted to create a more realistic depiction of the ousted president, Meta’s chatbot replied, “I can’t generate that.” Meta declined to comment. Alongside the popular tools created by tech giants, there is a growing assortment of lesser-known image generators from other A.I. companies. Many of those tools have few guardrails, allowing meme-makers and activists to create A.I. fakes targeting politicians, celebrities, and other notable people. In the Times’s tests, tools including Z-Image, Reve, and Seedream responded by creating realistic images of Maduro being arrested outside a military aircraft. That accessibility matters. Generative A.I. tools are more available than ever before, said Mariví Marín, director of ProboxVE, a nonprofit that analyzes digital ecosystems in Latin America. She wrote in a text that the Maduro deepfakes that emerged over the weekend, most of them before an official photo was released, were a response to “a collective need to fill in ‘the perfect image’ of a moment that each Venezuelan imagines somewhat differently,” a kind of “reaction to the urgency of filling that visual gap (with real or unreal content) as quickly as possible.” In other words, the fakes were not only a technical failure or a platform failure; they were also a human response to uncertainty. People wanted to see the moment, and when they could not, they made it—or shared someone else’s version of it.
The context in Venezuela made the confusion worse. The country is among the most restrictive in the world for press freedoms, with limited media sources, making trustworthy information hard to come by. In that news vacuum, the mistrust bred by deepfakes is exacerbated. Gutiérrez said many people were skeptical that the image of Maduro posted by Trump was authentic. “It’s funny, but very common: Doubt the truth and believe the lie,” he said. That inversion is one of the deepest dangers of synthetic media. It is not just that people may believe a fake image; it is that they may stop believing real ones. Latin American fact-checkers like Gutiérrez said they had been primed by earlier synthetic videos and images of Trump and former President Joseph R. Biden Jr. to be on the lookout for deepfakes of public figures. Since the pandemic, fact-checking groups across the region have collaborated to tackle cross-border rumors and conspiracy theories that surface during international conflicts and elections. Their work has become a kind of emergency response system for truth, but it is a system under constant strain. The Maduro episode showed how quickly that strain can become visible: fact-checkers had to identify fakes while official images were still scarce, while news outlets were already circulating suspect pictures, and while audiences were forming judgments in real time. The result was a familiar but painful paradox—the more convincing the fakes become, the more people may doubt everything, including the evidence that is real.
Gutiérrez summed up the difficulty with a candor that captures the mood of the moment: “It just took a lot of work, because we always lose the battle to convince people of the truth.” That line is not just about one weekend of fake Maduro photos. It is about the new normal of breaking news, where the first images people see may be invented, where official photographs may arrive late, and where platforms, regulators, and fact-checkers are often running behind. The Maduro case is a warning that the threat of A.I. fakes is no longer theoretical. The tools are fast, widely available, and capable of producing lifelike political imagery in seconds. Some companies have watermarks and rules, but the tests show that safeguards can be inconsistent or easy to route around. Lesser-known generators may have almost no guardrails at all. And in countries with restricted press freedoms or low institutional trust, the damage can be especially severe. The human stakes are clear: a person scrolling through a feed may not be doing forensic analysis; they may simply be trying to understand a dramatic event. When the visual record is polluted, that ordinary act of understanding becomes a minefield. The lesson is not that truth is impossible, but that it now requires more deliberate effort—stronger watermarking, faster platform enforcement, better media literacy, and sustained support for local fact-checkers. The Maduro deepfakes were a glimpse of a future that has already arrived: breaking news becomes a contest over the image, and the people trying to sort fact from fiction are left doing hard, urgent work while the rest of the world watches, shares, and sometimes believes the lie.